DynaWeave trains language agents to handle changing tasks dynamically
DynaWeave: Learning to Adapt through Dynamic Scenario Synthesis
Computation and Language
Summary
Artificial agents that use language models often face problems when unexpected changes happen in their environment, like sudden updates or interruptions from other users. The authors created DynaWeave, a tool that automatically turns fixed tasks into changing scenarios by adding external events that the agent must respond to. They tested this approach by gathering many example interactions and fine-tuning language models, showing the agents got much better at handling dynamic situations. This helps agents be more flexible and successful in real-world settings where things don’t always stay the same.
What this means in practice
- •For software developers building ai agents: Create more robust AI agents that adapt to changing task requirements and external events in interactive environments.
- •For automation engineers in smart services: Improve automated workflows by simulating unexpected external changes to test adaptability of tool-using agents.
Authors
Xuan Yang, Chunxiao Liu, Hongsheng Xin, Hao Xu, Ning Miao
Abstract
Large language model (LLM) agents increasingly use tools to complete tasks in stateful environments. However, many training environments remain largely static apart from changes caused by the agent's own actions, whereas external events introduced by other users, services, or sudden contingencies in real environments can invalidate ongoing plans. To address this gap, we introduce DynaWeave, an automated synthesis framework that transforms existing static tasks into verifiable dynamic scenarios. The framework generates exogenous events relevant to each task, covering changes in task requirements, environment state, and external information. It uses directed acyclic graphs (DAGs) to represent dependencies between events and an execution engine to activate them during agent interaction. A dynamic verifier revises only the evaluation checks affected by events that actually occur and retains all other checks. Using DynaWeave, we collect 2,095 verified interaction trajectories across 36 environments and use them for supervised fine-tuning of Qwen3.5 models. Experiments on Gaia2, CostBench, and ToolMaze show that DynaWeave achieves consistent improvements in overall performance across all three evaluated model sizes, increasing the task completion rate on CostBench by 67.34 percentage points and even exceeding the performance of Gemini-3.1-Pro-Preview on ToolMaze. These results support dynamic scenario synthesis as a practical approach to training agents to adapt to changing environments. Our code is available at https://github.com/Miaow-Lab/DynaWeave.